Re: none
- From: Maciej Stachowiak <mjs eazel com>
- To: Havoc Pennington <hp redhat com>
- Cc: Tim Janik <timj gtk org>, Gtk+ Developers <gtk-devel-list gnome org>
- Subject: Re: none
- Date: 13 Dec 2000 17:18:33 -0800
Havoc Pennington <hp redhat com> writes:
> Tim Janik <timj gtk org> writes:
> > what? how's GtkObject deprecated?
> >
>
> It has exactly one feature not found in GObject, which is the
> "destroy" signal. Our claim is that "destroy" is only useful for GUI
> objects, so "destroy" should really be in GtkWidget. So GtkObject is
> just legacy cruft.
>
> Granted, it's legacy cruft we are never going to remove, since it
> would be too painful. But newly-written user code should only be using
> GtkObject for the "destroy" signal, and never using it otherwise. So
> GtkObject is deprecated (i.e. not recommended that you use it), though
> not deprecated in the sense that we are going to remove it.
>
> You should never type a function that begins gtk_object_ anymore and
> you should never use the GTK_OBJECT() cast, because in all cases you
> should be using GObject or GtkWidget. The only legit way to use
> GtkObject is too connect to "destroy".
This is sadly broken. A "destroy" signal is needed for lots of non-GUI
objects. It gets used all the time in non-GUI Bonobo code, for
instance, so factory servers know when all their objects have gone
away.
Last time this came up Havoc or somebody said to just do
gobject_set_data with a destroy notification, but that seems
needlessly arcane. The "destroy" signal should be moved to
GObject. Even if it does not make sense for some non-gui objects,
there are cases where it's clearly needed.
Regards,
Maciej
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]